Facial Expression Recognition Using Convolutional Neural Network

Abstract

Facial expression conveys the emotional state of human beings. Facial expressions are a common form of non-verbal communication that helps to transfer necessary information or data from one person to another. However, in today’s world with increasing demand for artificial intelligence, recognition of facial expressions is a challenging task in solving problems related to artificial intelligence, machine learning, and computer vision. In this paper, we present an approach that helps to classify different types of facial expressions using Convolutional Neural Network (CNN) algorithm. The proposed model is a Neural Network architecture that is based on sharing of weights and optimizing parameters using CNN algorithm. Two Models are designed using this algorithm which is named Simple CNN and Improved CNN models having different convolution layers. Architecture designs of these two models are different from each other. The input of our system is grayscale images which consist of expressions of different faces. Using Input as grayscale images, both CNN models are trained and parameters optimized in neural network. Output of system is seven common facial expressions such as happy, anger, sad, surprise, fear, disgust, and neutral. To achieve better experimental results of designed model, graph of loss and accuracy is plotted for both of the above models. The overall simulation results prove that Improved CNN model improves the accuracy of facial expression recognition as compared to Simple CNN model. As result, analysis of confusion matrix is also obtained for Simple CNN and Improved CNN model.